Evaluation And Prediction Of Slope Stability Based On Engineering Fuzzy Set Theory

Landslide frequently occurs in China that threatens the people's life and property and economic construction greatly. With the rapid development of our modern construction, more and more high-steep and deep slopes appear in many engineering fields such as hydroelectricity, mine, harbor, road, railway and energy, etc. Therefore, how to analyze and evaluate slope stability reliably and effectively is very significant to slope safety and landslide forecast in theory and practice. This paper reviews the development actuality of slope stability research, indicates the shortcomings and development trend of various analytical methods and brings forward the research object and content. The results are obtained as follows:(1) The fuzzy evaluation of slope stability is a promising method. But many concerned literatures process concept and definition absolutely based on fuzzy set theory in slope stability analysis, it makes the determination of membership function so difficult and subjective that seriously barriers the application of this kind of methods in engineering. Therefore, the engineering fuzzy sets theory created by Prof. Chen Shouyu is introduced to slope stability analysis, and its application in slope stability evaluation is studied by this paper.(2) Many engineering cases with clustering characteristics provide abundant knowledge for slope stability analysis. The fuzzy clustering interactive computing model presented by Prof. Chen Shouyu is applied to the clustering analysis of many engineering cases, and its incomplete convergence is indicated. For solving the problem the fuzzy similar clustering model is put forward by this paper on basis of fuzzy optimal theory. The calculation results show that the fuzzy similar clustering model can class the clustering samples naturally and avoid above mentioned problem.(3) Case-based reasoning is a new artificial intelligent technology that developed in recent years. It provides a new feasible approach for slope stability evaluation. On the basis of the case-based reasoning idea the fuzzy optimization theory is applied to the comparison of the slope similarity, and the method of fuzzy optimal selection of similar slopes is put forward to analyze slope stability. The method can take many uncertain influencing factors into account, obtain characteristics and knowledge from the known engineering cases and find the best similar slope sample to the object slope that needs to be evaluated its stability by comparing their relative membership degrees. In this way the stability of the object slope can be evaluated by that of the best similar slope sample. The practical calculation indicates that the method is effective in slope stability evaluation.(4) In the fuzzy evaluation of slope stability the index weight is generally determined by the knowledge, experience and wisdom of the analyzer. It is so subjective and random and insufficient to the slope stability evaluation influenced by numerous factors that it is necessary to determine the objective weight using theobjective information hidden in the known decision matrix. However, the preference of the analyzer is then neglected in the determination of the so-called objective weight. So it is urgent to find a decision model that can comprehensively take the objectiveness of the decision matrix and the subjective motivity of the analyzer into account so that the slope stability can be evaluated correctly. Therefore, the multi-pole fuzzy pattern recognition model of subjective and objective weight determination presented by Prof. Chen Shouyu is applied to the slope stability analysis, and good result is achieved.(5) In view of the difficulty of weight determination of many influencing factors in the fuzzy evaluation of slope stability, the application of the multi-layer fuzzy pattern recognition theory and model in slope stability evaluation is studied. The whole appraisal system is composed of three layers. As input layer the first layer consists of all influencing factors. The second layer has some subsystems, of which every one consists of some similar factors. The third layer has only one subsystem that is output layer. The weights of influencing factors of every subsystem and the weights of subsystems can be determined respectively by comparing the importance of factors of every subsystem and that of the subsystems. In this way the weight determination is relatively easier and more reasonable than that by directly comparing all factors. The outputs of the first and second layers are both calculated by duel-pole fuzzy pattern recognition models, and that of the third is by multi-pole fuzzy pattern recognition model. The engineering case analysis shows that the application of this model in slope stability evaluation can achieve good result.(6) The evolving process of landslide is so complicated that some analytical methods are often used together in slope stability analysis and prediction. Here some common displacement-time statistical models are discussed. It is found firstly that the restored creep curve for slope deformation description has the normal distribution. Then the displacement-time normal and semi-normal models, the corresponding time - displacement logarithm models and their coupling models are successively proposed. The comparison of the presented models with the typical multivariable nonlinear regression model and the animal increment curve model shows that the forecast made by the presented models to the given case is satisfying. Furthermore, the fuzziness of the slope engineering is considered. The weight determination in the model calculation is studied using the fuzzy optimization theory, and the practical mathematical model for the weight determination of the data samples is given.(7) "With the presentation of the numerous landslide displacement models, it is necessary to build an appropriate criterion or method to appraise the known models so that the optimal evaluation model can be chosen. Today the common used method is the model evaluation using the statistical characteristics. In fact, the match of any model and the measurement is relative that is random and fuzzy.Therefore, the fuzzy optimization theory is applied to the selection of landslide displacement models, and a fuzzy optimal selection method of similar landslide displacement curves is put forward. The use of the method is very simple that provide an effective approach for the selection of the landslide displacement models.(8) Being widely used in the evaluation of the slope stability BP neural network has bad convergence and local minimum problem. Comparatively RBF artificial neural network (RBF ANN) is better than BP in approximation, classification and learning speed. At the same time, the common used k-mean algorithm of RBF ANN has some problems to be solved. For this reason, the RBF ANN, the fuzzy optimization theory and the fuzzy clustering analysis theory are considered together, and two new intelligent models for slope stability evaluation and prediction are put forward: the fuzzy similar clustering RBF ANN and the fuzzy similar clustering recognition ANN. The practical calculation shows that good results can be obtained in slope stability evaluation and prediction by the two models.Lastly, the whole paper is summarized, and some promising problems are discussed.